Research On Parameters Estimation And Modulation Classification Of Communication Signal

The paper discusses how to design a signal processing platform which can implement modulation recognition and parameters estimation on multiple common digital signals.Based on analysis and research on the spectral-correlation and other characters of ASK,FSK,MSK,MPSK(BPSK,QPSK,8PSK),MQAM (16QAM,32QAM,64QAM),the paper first designs the algorithm of modulation recognition and parameters estimation. And then, the processing project of the whole system is proposed. Finally, the signal processing platform is designed by DSP+FPGA processing. The main contents of this paper are as follows:Firstly, based on analysis and research on the signal pre-processing, the algorithm of multi-peak separation of the spectrum is designed, and it shows good performance through simulation.Meanwhile, the rough estimation algorithm of baud rate of ASK and FSK signals is designed, and the MSE of estimation is below 10-1 higher than 2dB SNR.Secondly, the comparison between the algorithm of periodogram and spectrum barycenter is studied. And then, the modified spectrum barycenter algorithm is presented based on system requirement. The good performance in lower SNR of the modified spectrum barycenter algorithm has been proved by the simulation results.Thirdly, symbol rate estimation based on cyclic spectrum is studied. The algorithm named f-f0 section method which is used for estimating the symbol rate of MPSK and MQAM signals and the algorithm namedf= 0 section method which is used for estimating the symbol rate of MSK signal have been presented. When it comes to the signal of ASK and FSK, a novel algorithm named window-added method which is based on the f= f0 section method has been proposed. Simulation results show that both of the algorithms have good performance in lower SNR.And then, based on the spectral-correlation characters analysis of the five kinds of signals above, a series of cyclic spectrum features which are used for modulation recognition have been extracted. However, for the problems of differentiating ASK signal from one spectrum peak of FSK signal and MPSK signal from MQAM signal and MQAM subclass, the method based on spectral-correlation characters can not solve them perfectly. And these are difficulties of this paper. By designing the algorithm based on frequency correlation feature, second-order amplitude moment and second-order amplitude statistic moment, the paper has solved the difficulties. The good performance in lower SNR of these features has been proved by the simulation results. In 5dB SNR, the average recognition rate can reach to 95%, while the lowest rate can reach to 90%.Finally, the feasible system project based on the researches above has been designed, which balances between the performance and the complexity. using the DSP chip of TMS320VC6416 and the FPGA chip of EP2S90F1020C5, the implementation and debugging of all modules of the system have been completed, and the joint debugging of DSP and FPGA have also been accomplished. Finally, the base band testing report has been given. It's shown that the unit of signal processing can commendably complete parameter estimation and modulation recognition in lower SNR. Further more, both the estimation precision and recognition rate can satisfy the requirement of the system.